View Single Post
Old 25th June 2020, 02:06 PM   #10
The Clarity Is Devastating
Myriad's Avatar
Join Date: Nov 2006
Location: Betwixt
Posts: 18,255
Originally Posted by WhatRoughBeast View Post
So here's a new one (at least, it is to me, and I can't seem to find anything about it on the web).

I went to Worldometer and looked at global COVID deaths for the last 63 days. Totalling global deaths by day of the week produces

Monday - 30,717
Tuesday - 33,106
Wednesday - 47,900
Thursday - 46,165
Friday - 48,629
Saturday - 46,860
Sunday - 41,115

The US follows a similar pattern over the last 63 days

Monday - 6,153
Tuesday - 7,391
Wednesday - 13,178
Thursday - 13,307
Friday - 13,612
Saturday - 10,548
Sunday - 10,454

Looking at the last 28 days for the US, thinking that we've clearly gotten better at keeping deaths down and this might have some effect,

Monday - 1,610
Tuesday - 2,105
Wednesday - 3,941
Thursday - 3,683
Friday - 3,908
Saturday - 3,698
Sunday - 2,998

and if anything the Monday/Tuesday effect has become more pronounced

I like to think of myself as pretty good at coming up with explanations for weird behavior, but this has got me stumped. Two days out of the week have death rates about 30% below the other five days. What's so special about Monday and Tuesday that people are not dying?

Any suggestions?

All the statistics sites list "deaths reported that day." "Reported" means reaching the final stage of the information being received and listed by the relevant state agency and then getting passed on to whomever's posting the graphs. The weekly patterns relate mostly to which hospital departments and public health departments and intermediate state agencies are closed on weekend days, meaning not many of them reach the final tally point on Sun. or Mon.

In their internal reports, Massachusetts back-dates each death to the date it actually occurred, and also back-dates each new case (new positive test result) to the date the swab was taken. That makes the time series a lot smoother (though there's still a lull in new cases each Sun-Mon because fewer tests are actually performed on those days). But the Massachusetts data you see at Google or Worldometer doesn't show the back-dating. They want one number per measure per day (new deaths reported, new cases reported) and don't want to have to change the numbers from previous days. So you're seeing the crude sloppy data, and you can't easily discern trends on time scales shorter than weekly.
A zÝmbie once bit my sister...
Myriad is offline   Quote this post in a PM   Nominate this post for this month's language award Copy a direct link to this post Back to Top